How does the brain predict and construct our perception of the world? Is consciousness dependent on active brain cells, or could inactive neurons also contribute?
How does consciousness arise from the activity of our brains? Do we need active brain cells to be conscious, or can inactive neurons play a role?
How does our brain create our sense of space, for instance around blind spots in the visual field of an individual eye? Do we need to actively explore our environment to be conscious of it, or can sensory activity result in conscious experience without motor activity? The answers to these questions could have profound implications not just for our understanding of consciousness, but for medicine – from helping doctors better assess consciousness levels in patients who are unable to communicate to improving treatments for conditions ranging from stroke to amyotrophic lateral sclerosis (ALS).
Led by Cyriel Pennartz, Professor of Cognitive and Systems Neuroscience at the University of Amsterdam, Giulio Tononi, Professor of Psychiatry, Distinguished Professor in Consciousness Science, and Karl Friston, Professor of Imaging Neuroscience and Wellcome Principal Research Fellow at University College London, an international team of scientists and researchers have designed a series of experiments that could not only answer some of these questions, but potentially advance our understanding of consciousness.
The project, contrasting predictions of the Integrated Information Theory (IIT) and Predictive Processing accounts of consciousness — or INTREPID — is one of five Structured Adversarial Collaborations within the Accelerating Research on Consciousness initiative being funded by the Templeton World Charity Foundation (TWCF). It originated from a fundamental disagreement between theories about how consciousness arises in the brain. As Pennartz explains, one theoretical framework, Predictive Processing, builds on the idea that what we perceive is not the world directly but rather an internally generated appearance constructed by our brain to make sense of the imperfect trains of electrical impulses from our sense organs. The brain constructs this appearance in a continuous balancing act between predictions of its sensory input, based on its prior knowledge, and the actual sensory input, and the least surprising input then determines what is consciously perceived.
More specifically, the project contrasts two Predictive Processing theories. The Predictive Processing - Active Inference (PP-AI) theory proposes that consciousness requires active exploration of our environment, such as shifting attention or moving our eyes to look at an object; while the Predictive Processing - Neurorepresentationalism (PP-NREP) theory suggests that consciousness arises from the brain's ongoing predictions about sensory input, but does not require overt or covert actions like eye movements or directed attention.
“Active inference basically says your consciousness starts with an action towards the world. And then the world answers back to that by giving you new sensory information”, says Pennartz. “For instance, you make a reach movement towards an object. And then the sensors in your hand, your fingers, would say, OK, this does indeed feel like an apple or a ball or whatever.”
“For NREP, the inference or the generation of predictive representations is primarily based on information from multiple senses, whereas body activity is not required per se for consciousness. So, based on the sensory inputs my brain gets, I predict what's most likely going on now in the world. Body movement or attention can play a role in this, but do not have to be present. NREP then proposes a scheme that builds high-level, conscious representations (or predictions) from lower level processes, eventually going down to the level of single brain cells”.
Predictive processing, developed by Karl Friston, is based on a general theory about brain function. Starting from the general account, the project (which is a group effort) narrows this down to the two accounts of consciousness. In contrast, the Integrated Information Theory (IIT), developed by Giulio Tononi, starts from the consideration of what subjective experience is and then proposes that consciousness emerges from the structural organization of potential causes and effects within neural networks even when neurons are not actively firing. This leads to some controversial implications, including the idea that rudimentary consciousness could exist in systems without active neural firing, and potentially even in non-biological systems like weather patterns.
“IIT is an intriguing theory, which offers a quite beautiful account of the hard-to-express richness of conscious experience, but it is also challenging to understand,” says Jakob Hohwy, Professor of Philosophy at the Centre for Consciousness and Contemplative Studies at Monash University, whose lab is testing PP-AI. “This adversarial collaboration makes important technical steps to test IIT empirically, and it has been an interesting intellectual journey for predictive processing theories to make our own predictions about what will happen in these tests.”
The contrasting positions of the three theories give rise to testable predictions. Specifically, the principle of adversarial collaboration will be borne out by a complementary hypothesis-testing across three main experiments, each of which is designed in a way that will be able to test incompatible predictions of at least two groups of theories (e.g., IIT vs. both PP-NREP and PP-AI). Together, the experimental results should provide evidence to support one theory while challenging the other two.
"What makes this project special is that it's not just scientists trying to prove their own theory right," says Pennartz. "Instead, we have advocates of different theories working together to design experiments that will truly put their ideas to the test."
To test these competing ideas, the researchers have designed the following main experiments:
In addition to advancing our understanding of consciousness, resolving these theoretical differences could – in the near future – have practical applications, like helping to develop new treatments for stroke patients with visual field defects or spatial awareness problems, or for ALS patients. As Pennartz notes, researchers have already developed brain-computer interfaces that can help paralyzed ALS patients communicate.
"We can read from the brain signals the translation on the computer screen of what they want to say," he explains. “Understanding consciousness better could help improve these technologies, as well as help clinicians make more informed decisions about patient care and recovery prospects.”
To ensure reliable results, each experiment will be independently replicated in a second laboratory to validate findings. Researchers will also use advanced brain imaging techniques to track neural activity and confirm their findings.
“In our test of PP-AI we use meticulous vision science and advanced EEG techniques, and I am amazed at the technical innovations in the other experiments, pushing the limits of controlling and taking measurements from small populations of neurons during ongoing tasks”, says Hohwy.
Pennartz adds: “In the experiment on the role of inactive neurons, Dr. Umberto Olcese and colleagues are combining refined illumination of the brain of mice performing a spatial vision task with optogenetics, which uses light to change the activity state of genetically modified brain cells. This is at the limit of what is currently possible in neuroscience”.
The project's governance structure is rigorous, and includes a steering committee, management boards for each experiment, and an advisory board featuring prominent researchers in the field. This careful oversight aims to ensure the research maintains high scientific standards, including that the findings are reproducible, while fostering collaboration between different theoretical perspectives.
While ambitious, the researchers acknowledge that even these carefully designed experiments may not definitively prove or disprove any single theory. However, they expect the results will significantly advance our understanding of consciousness and help refine existing theories. As Hohwy says “The key is that every theorist has made predictions about all the experiments and in this way, the adversarial collaboration is a race for evidence, where every theory has a lot of skin in the game”. Indeed, the theorists are not allowed to change their predictions after the experiment has started.
"Understanding consciousness is one of the greatest challenges in science," notes Professor Pennartz. "This project may not give us all the answers, but it represents an important step forward in how we scientifically study consciousness."
The project, which will run for several years, plans to share results as they emerge with both the scientific community and the public – all data and analysis methods will be made available, allowing other researchers to verify and build upon the findings. Additionally, a website – https://acceleratingresearch.org – will be continually updated with information about this and four other similar projects being led and funded by TWCF.
As this research continues, it may not only help resolve long-standing debates about consciousness but also provide new hope for patients with neurological and psychiatric conditions, as well as conditions affecting conscious experience, demonstrating how theoretical neuroscience can bridge the gap between philosophical questions about consciousness and practical medical treatments.
To better understand existing theories of consciousness, Templeton World Charity Foundation (TWCF) funds a series of 5 ambitious research projects, known as Structured Adversarial Collaborations, to allow researchers to evaluate competing hypotheses through carefully designed experiments and rigorous cross-laboratory validation.
This blog is part of a series that complements the launch of Accelerating Research, a unique platform developed by Dawid Potgieter and DataCite with TWCF funding, showcasing results from each Structured Adversarial Collaboration project as it progresses.
